DATA COLLECTION

Single-item measure of food insecurity used in

the National Health Survey may underestimate
prevalence in Australia
Rebecca McKechnie,1 Gavin Turrell,2,3 Katrina Giskes,1 Danielle Gallegos1

F
ood and nutrition security is defined by
Abstract
the Food and Agricultural Organization
(2012) as: “when all people at all times Objectives: To compare prevalence estimates of food insecurity using a single-item measure,
have physical, social and economic access to with three adaptations of the United States Department of Agriculture Food Security Survey
food, which is safe and consumed in sufficient Module (USDA-FSSM).
quantity and quality to meet their dietary Methods: Data were collected by postal survey, from individuals aged ≥18 years from
needs and food preferences, and is supported disadvantaged suburbs of Brisbane, Australia (n= 505, 53% response). Food security status was
by an environment of adequate sanitation, ascertained by the Australian single-item measure, and the 6-, 10- and 18-item versions of the
health services and care, allowing for a USDA-FSSM. Prevalence estimates of food insecurity and different levels of severity of food
healthy and active life”.1 insecurity estimated by each tool were determined. Data were analysed using McNemar’s test,
Food and nutrition security is underpinned polychoric correlation and Rasch analyses.
by four dimensions, being: i) availability of Results: The prevalence of food insecurity was 19.5% using the single-item measure;
sufficient amounts of nutritionally adequate significantly less than the 24.4%, 22.8% and 21.1% identified using the 18-item, 10-item and
food; ii) adequate financial and physical 6-item versions of the USDA-FSSM, respectively. Rasch analyses revealed that overall the USDA-
access to such foods; iii) the infrastructure, FSSM may be a valid tool for the measurement of food insecurity within the current sample.
resources and ability to utilise foods; and,
Conclusion: The measure of food insecurity employed in national surveys in Australia may
iv) stability of the previous three pillars over
underestimate its prevalence and public health significance.
time. Any absence or disruption of one or
more of these pillars results in food insecurity. Implications for public health: Future monitoring and surveillance efforts should seek to
employ a more accurate measure as the first step in recognising the right to food for all
Food insecurity may be associated with a
Australians.
range of adverse health outcomes. Among
adults, it may lead to dietary choices less Key words: food security, measurement, monitoring and surveillance
consistent with recommendations for good
health,2,3 overweight or obesity,4,5 poorer The most commonly used measure to assess levels of severity (low food security, very low
self-reported general health6,7 and increased food insecurity at the household level is food security and low food security among
risk of chronic disease (such as heart disease, the 18-item United States Department of children). The measure is also able to be
diabetes and depression).8,9 Among children, Agriculture Food Security Survey Module shortened to a 10-item and 6-item sub-scale
food insecurity may result in poorer self- or (USDA-FSSM);16,17 this scale is used in national that can differentiate between low levels
parent-reported general health,10,11 lower surveys across high-income countries, such of food security and very low levels of food
academic achievement,12,13 behavioural and as the United States and Canada, and is the security. Adaptations of the USDA-FSSM,
conduct problems12 and poor health and social predominant measurement tool mentioned such as the 15-item Latin American and
outcomes in adulthood.14,15 These implications in literature. The USDA-FSSM comprises 18 Caribbean household food security scale
are of notable significance to public health, items that specifically assess financial access (ELCSA) tool and the 9-item Household Food
making food insecurity an issue worthy of to food; responses to the FSSM items are Insecurity Access Scale (HFIAS) also exist for
continued monitoring and surveillance. summed to allow for identification of varying use in less industrialised countries.18 In New

1. School of Exercise and Nutrition Sciences, Queensland University of Technology

2. Institute for Health and Ageing, Australian Catholic University, Victoria
3. School of Public Health and Social Work, Queensland University of Technology
Correspondence to: Professor Danielle Gallegos, School of Exercise and Nutrition Sciences, Queensland University of Technology, Victoria Park Rd, Kelvin Grove, QLD 4059;
e-mail: danielle.gallegos@qut.edu.au
Submitted: January 2018; Revision requested: April 2018; Accepted: May 2018
The authors have stated they have no conflict of interest.
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium,
provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
Aust NZ J Public Health. 2018; 42:389-95; doi: 10.1111/1753-6405.12812

2018 vol. 42 no. 4 Australian and New Zealand Journal of Public Health 389
© 2018 The Authors
McKechnie et al. Article

Zealand, food security status is assessed via a human subjects were approved by the 20 and 59 years. Individuals residing in
multi-item tool comprised of eight questions Queensland University of Technology Human suburbs that contained the required CCDs
included in the yearly National Health Survey. Research Ethics Committee (Project number were extracted and addresses were then
This scale has been validated for use among 0800000735). Written informed consent was geocoded to CCDs using MapInfo (version
the New Zealand population; however, obtained from all subjects. 11.5, MapInfo Corporation, Troy, New York).
its ability to be compared to international Finally, households within the required CCDs
data collected via other measures of food Study design were identified and 1,000 households were
security remains undetermined.19 These tools A cross-sectional design was employed to randomly selected to participate in the study.
primarily assess financial access to food. compare the single-item Australian measure
In Australia, a single-item measure of food to the 18-, 10- and 6-item versions of the Participants and data collection
insecurity that asks: “In the last 12 months USDA-FSSM. Data collection occurred via postal-
was there any time you have run out of food survey between March and May 2009.
and not been able to purchase more?” is Study scope and sampling The questionnaire was 12 pages in length,
incorporated in the three-yearly National The current study was conducted in the comprised primarily of previously validated
Health Survey (NHS) and, incorrectly, is used Brisbane Statistical Sub-Division (SSD) and items, and sought information on dietary
as an indicator for the spectrum of severity comprised 1,000 individuals randomly and health factors, household food security
of food insecurity.20 Unfortunately, food selected from households in the 80 most status and a range of socio-demographic
insecurity was not assessed during the recent socioeconomically disadvantaged census characteristics such as gender, age, country
National Health Survey in 2014–2015.21 collector districts (CCDs), being the most of birth, household income, family type and
The most recent data available from the disadvantaged 5% of CCDs in the SSD. CCDs Indigenous status.
2011–2012 Australian Health Survey (AHS) are the smallest units for which statistical data
(a larger survey that incorporates the are available in Australia.24 Food security status
National Health Survey as well as a more Sample size was determined with Food security was assessed using four
comprehensive diet and exercise survey, consideration given to objectives of other measures. Firstly, the single NHS question
the National Nutrition and Physical Activity aims of the study (published).25,26 Acceptable was included, which asked participants: “In
Survey), identified the prevalence of food values for meaningful and/or practically the previous 12 months was there any time
insecurity among the general Australian significant differences between variables you have run out of food and not been able
population to be approximately 5%.22 of interest (based on the aforementioned to purchase more?” The response option was
However, research has suggested that previous objectives) for food secure ‘yes/no’, and households were coded as either
the single-item may underestimate the compared to food insecure households food secure (negative response) or food
prevalence of food insecurity compared to were determined using results from existing insecure (affirmative response).
the more comprehensive USDA-FSSM.23 To literature. Starting figures determined using The 18-item USDA-FSSM was also
ascertain the true burden of food insecurity, standard calculations were multiplied by incorporated in the questionnaire, and
be able to make meaningful comparisons factors of 1.60 to account for an anticipated included questions relating to the food
between countries and guide policy and 60% non-response, 2.00 (the number of situation of the household in the previous
interventions to address this issue, it is groups for comparison) and 2.33 to account 12 months, such as running out of food
imperative that accurate measures of food for an uneven sample fraction (approximately and being unable to purchase more due to
insecurity be employed. 30% of the target population were expected financial constraints, being unable to afford
The current study compared the prevalence to be food insecure). The variable that yielded balanced meals, reducing the size of meals
of food insecurity in disadvantaged suburbs the largest required sample was selected for or skipping meals because of being unable
of Brisbane city in 2009, using different determination of the final sample size for the to afford food, and going hungry due to not
measures to ascertain food security status: overall research project; 1,000 participants being able to afford food. Similar questions
the single-item used in the NHS and the 18-, were required for 90% power, with a type one were asked in relation to children’s dietary
10- and 6-item iterations of the USDA-FSSM. error of 5%. behaviours (if children were present in
Furthermore, the psychometric properties The sampling of individuals was undertaken the household). Responses to these items
of the 18-, 10- and 6-item versions of the as follows. Firstly, the lowest 5% of CCDs were were summed (score of 1 for an affirmative
USDA-FSSM were assessed to identify the identified according to each CCD’s Index response and 0 for a negative response) and
appropriateness of the scales for a low- of Relative Socioeconomic Disadvantage scores were used to categorise households as
socioeconomic Australian sample and (IRSD) score, which was obtained from either: food secure (0 to 2); low food security
potential improvements to improve the the Australian Bureau of Statistics (ABS).27 (3 to 5 among households without children,
validity of these scales among an Australian Secondly, data providing name, gender, 3 to 7 among households with children); very
population. age and address were accessed via the low food security (6 to 10 among households
electoral roll; individuals between the ages without children, 8 to 12 among households
with children); and very low food security
Methods of 25 and 45 years were selected to capture
the age group with the greatest likelihood among children (13 to 18 among households
This study was conducted according to the of having a residential child (to address a with children).16 For the purpose of data
guidelines laid down in the Declaration related aim of the study); however, the final analyses, categories were dichotomised
of Helsinki and all procedures involving age range of respondents ranged between to ‘food secure’ or ‘food insecure’; these

390 Australian and New Zealand Journal of Public Health 2018 vol. 42 no. 4
© 2018 The Authors
Data Collection Underestimation of food insecurity in Australia

classifications were derived by initially McNemar’s test (cross tabulation) was Table 1: Demographic and food security
classifying ‘adult food insecurity’ (score of implemented to investigate the difference in characteristics of sample compared with the
three or higher based on the 10 adult-related prevalence estimated by each food security population of residents aged 20 – 59 years in the
items) and ‘child food insecurity’ (score of two screening tool as well as the 18-, 10- and 6- selected CCDs.
or higher based on the eight child-related item USDA tools with the raw score threshold Study Total resident
items). Households were then classified adjusted to allow for the identification of sample population
as food insecure if they were identified as the same prevalence as the single-item NHS (%) of selected
experiencing ‘adult food insecurity’, ‘child food measure. Finally, reliability analysis was used N = 487 CCDsa
insecurity’ or both. This scoring protocol is in to investigate internal reliability of the 18- (%)
line with recommendations by Nord,28 and item, 10-item and 6-item measures. Gender
the current scoring protocol implemented Male 44.5 48.5
in national surveys across Canada, as it Female 55.5 51.5
accounts for distortion in measurement when
Results Age group
20 – 29 years 16.8 27.3
comparing households with and without Of 1,000 individuals sampled, three were
30 – 39 years 30.0 26.4
children.28 unable to speak English, 49 no longer resided 40 – 49 years 35.1 25.1
The USDA-FSSM can be shortened to a 10- at the address listed on the electoral roll 50 – 59 years 18.1 21.2
item and 6-item version.16 The appropriate and one was overseas, leaving 947 potential Country of birth
questions were isolated and scored; participants. A total of 505 completed Oceanic/ Antarctica 74.4 75.3
households were categorised as either: food questionnaires were returned, resulting in Europe 7.4 8.6
secure, low food security or very low food a final response rate of 53%, 487 for which Africa & Middle East 2.9 1.4
security, and data dichotomised to ‘food complete data regarding food security status Asia 13.8 13.3
secure’ and ‘food insecure’ for the purpose of was available. Americas 1.5 1.4
analyses (with food insecurity status assigned Household Income
Table 1 summarises the demographic
to households with a score of three or higher ≤ $499 22.9 23.6
characteristics of participants in the study
$500–$799 17.7 25
when using the 10-item iteration, and two compared to the characteristics of the general
$800–$999 11.5 8.7
or higher when using the 6-item iteration, population in the CCDs sampled. The sample $1000–$1399 16.2 22.5
respectively).16 was comparable to the general population $1400–$2499 27.1 16.4
residing in the selected CCDs in terms of ≥ $2500 5.6 3.7
Analyses
gender, country of birth and Indigenous Family type
Data were analysed using jMetrik freeware, status; however, the sample included a Couple, no children 25.1 29.8
SPSS version 16 (SPSS Inc., Chicago, IL, USA) slightly greater proportion of higher income Couple, with children 49.2 37.4
and R software. households than the population within the Single-parent family 25.7 32.8
First, jMetrik freeware was used to selected CCDs. Indigenous status
undertake Rasch analyses to investigate the Non-indigenous 95.7 94.1
psychometric characteristics of the items Psychometric properties of Aboriginal 3.5 5.1
Torres Strait Island 0.5 0.4
in each of the USDA-FSSM adaptations. 18-, 10- and 6-item USDA-FSSM
Aboriginal & Torres Strait Island 0.3 0.4
INFIT values were used to identify the Results from the Rasch analysis are Note:
discriminative capacity of each item by summarised in Table 2. Of the full 18 items a: Data from the 2006 Australian census for CCDs with the 5% lowest
assessing the expected responses compared in the USDA-FSSM, 10 fell within the widely
IRSD scores in the Brisbane Statistical Subdivision.
to actual responses. INFIT values investigate accepted range for fit and four fell within
whether items in a scale measure the same between the 18-, 10- and 6-item USDA-FSSM
the wider acceptable limits, leaving only
construct independently of each other. and the single item from the NHS (r = 0.948,
four that fell outside both the standard and
When the responses to an item fit the 0.972 and 0.948, respectively).
wider acceptable ranges. Of these, two were
model perfectly, an INFIT value of one is redundant and two appeared to have been
generated. INFIT values below one indicate provided more affirmative responses than
Differences in prevalence between
fewer affirmative responses than expected, predicted by the model. In the 10-item short single-item NHS measure and USDA-
suggesting an item is redundant. INFIT values form US adaptation, two fell outside the wider FSSM adaptations
above one suggest that that more people limits of agreement (one redundant and one The prevalence of food insecurity estimated
answered affirmatively than predicted by the overvalued). Finally of the 6-item version, by each measure is summarised in Table 3.
model, and that the information contributed three fell outside the standard acceptable The 18-item USDA-FSSM identified a greater
to by the item may be ‘overvalued’. Based on range; however, only one of these fell outside proportion of food insecure households
widely accepted ranges, cut-off values were the wider limits of agreement. compared to the 10- and 6-item USDA-FSSM
set at a recommended range of 0.8 to 1.2,
and the single question used in the Australian
with a wider acceptable range of 0.7–1.3.29
Correlation between single-item NHS NHS (24.4% versus 22.8%, 22.1% and 19.5%,
Second, using SPSS and R, correlations measure and USDA-FSSM adaptations respectively). McNemar’s test showed that
between each of the USDA-FSSM adaptations the 4.9%, 3.3% and 2.6% differences (absolute
Findings from the polychoric correlation
and the single-item NHS measure were difference) between the single-item NHS
analysis revealed a high level of correlation
investigated via polychoric correlation.

2018 vol. 42 no. 4 Australian and New Zealand Journal of Public Health 391
© 2018 The Authors
McKechnie et al. Article

question and the 18-item (p<0.001), 10-item from the NHS is four for the full 18 items (one 10.9%, 11% and 8.6% reported very low
and 6-item (p=0.01) tools, respectively, were point below the current threshold for food food security, respectively. The 18-item tool
significant. Reliability analyses showed the insecurity), and three and two for the 10- and identified that 0.8% of households reported
6-item, 10-item and 18-item measures to have 6-item USDA-FSSM respectively, (equal to low food security among children; however,
comparable reliabilities (Cronbach’s alpha the current thresholds used to identify food the 10- and 6-item versions were not able
0.89 for each). insecurity using these tools). to differentiate this level of severity of food
Table 4 provides a summary of the cumulative insecurity.
per cent for the raw score of each of the Food insecurity severity as estimated
USDA-FSSM adaptations. Based on these, by each measure Discussion
the threshold score at which the USDA-FSSM The 18-item, 10-item and 6-item tools
most closely replicates the proportion of food estimated that 12.7%, 11.8% and 13.6% of The study showed a high prevalence of
insecurity as identified by the single-item the sample reported low food security and food insecurity among disadvantaged
urban areas of Brisbane. The food security
screening question used in the NHS was
Table 2: Item fit statistics for 18-item, 10-item and 6-item United States Department of Agriculture (USDA) Food
highly correlated with each of the USDA-
Security Survey Module (FSSM) among an urbanised disadvantaged sample in Brisbane, Australia.
FSSM adaptations; however, it provided the
Item 18-item 10-item 6-item
USDA-FSSM USDA-FSSM USDA-FSSM
lowest estimate of food insecurity (19.5%)
Adult items compared to the three versions of the
Worried whether food would run out 1.33b 1.42b NA USDA-FSSM (24.4%, 23% and 21% for the 18-,
Ran out of food 0.97 0.98 1.24a 10- and 6-item USDA-FSSM, respectively).
Couldn’t afford healthy meals 1.06 1.11 1.26a The NHS question was also unable to
Adults cut size or skip meals 0.65b 0.63b 0.61b differentiate between the levels of severity
Frequency adults cut/ skipped meals 0.73 0.87 0.98 of food insecurity. These findings suggest
Ate less than thought should 0.65b 0.73a 0.72 that the current national estimates of food
Hungry but didn’t eat 0.81 0.89 0.98 insecurity may potentially underestimate the
Lose weight 0.87 1.21a NA prevalence and public health significance
Adults not eat for whole day 0.78a 0.84 NA of food insecurity and its experience among
Frequency not eat whole day 0.84 1.00 NA various socio-demographic groups.
Child items Our findings are consistent with previous
Relied on few kinds foods to feed children 1.61b NA NA studies, which found that the single question
Couldn’t feed children health meal 1.24a NA NA used in the Australian NHS underestimated
Children not eating enough 0.98 NA NA the prevalence of food insecurity by
Cut size children’s meals 0.88 NA NA approximately five percentage points23 and
Children hungry 0.78a NA NA
that the 6-item measure underestimated
Children skipped meals 1.90b NA NA
food insecurity by up to eight percentage
Frequency children skipped meals 1.17 NA NA
points when compared to the 18-item
Children not eat for whole day 1.17 NA NA
USDA-FSSM.30 Food insecurity is a complex
Notes:
a: INFIT value falls outside standard acceptable limit of 0.8–1.2
issue, with varying levels of severity; as such,
b: INFIT value falls outside wider acceptable limit of 0.7–1.3 it is unlikely to be adequately captured by a
single-item question. The current estimate
of food insecurity in the general Australian
Table 3: Prevalence and level of severity of food insecurity in the sample as measured by four different
population is approximately 5% and is based
measurement tools (N = 487).
on the single item used in the NHS.22 At the
Single-item 6-item 10-item 18-item
Australian USDA-FSSMa USDA-FSSMa USDA-FSSM population level, a difference of 5% between
questiona,b,c the full 18-item USDA-FSSM scale and the
Prevalence of food insecurity single NHS item may equate to an extra
Food secure (%) 78.4 75.4 74.7 72.7 300,000 households that are not identified by
Food insecure (%) 19.5 22.1 22.8 24.4 the single question. Underestimation by this
Missing (%) 2.1 2.5 2.5 2.9 tool may suggest that the true prevalence
Level of severity of food insecurity of food insecurity among the Australian
Food secure (%) 78.4 75.4 74.7 72.7 population may be closer to 10%, and the
Low food security (%) NA 13.6 11.8 12.7 findings of the current study show that the
Very low food security (%) NA 8.6 11.0 10.9 prevalence of food insecurity is markedly
Very low food security among children (%) NA NA NA 0.8 higher in disadvantaged urban areas and
Notes: may range from one-in-four to one-in-five
USDA-FSSM, United States Department of Agriculture Food Security Survey Module.
households.
a: Significantly different to 18-item USDA-FSSM
b: Significantly different to 10-item USDA-FSSM Identification of the threshold at which the
c: Significantly different to 6-item USDA-FSSM prevalence of food insecurity estimated by

392 Australian and New Zealand Journal of Public Health 2018 vol. 42 no. 4
© 2018 The Authors
Data Collection Underestimation of food insecurity in Australia

each of the USDA-FSSM adaptations most and specific sub-groups of interest may be Table 4: Cumulative percentage of affirmative
closely matched the prevalence identified warranted, to further adapt the USDA-FSSM to responses by raw score for 18-item, 10-item and
by the single NHS question revealed that suit the Australian population. 6-item United States Department of Agriculture
the single-item accurately captures those Overall, the findings of the current study (USDA) Food Security Survey Module (FSSM) and
from households experiencing very low suggest that using the 18-item measure threshold at which % food insecure meets %
food security and very low food security potentially provides a more accurate and identified by single-item National Health Survey
among children. The item, however, does descriptive account of food insecurity; (NHS) measure.
not effectively capture all households that this scale is a reasonably valid tool Raw 18-Item 10-item 6-item
experiencing low food security, resulting for measurement in the current sample score USDA-FSSM USDA-FSSM USDA-FSSM
in the underestimation identified above. If (disadvantaged households in Brisbane); (%) (%) (%)
the single-item measure continues to be and that it may be a valid option for other 0 49.3 51.7 64.5
used for monitoring and surveillance, this population sub-groups, as well as the 1 63.9 65.7 75.4
should be acknowledged as a limitation of broader Australian population. However, in 2 71.9 74.7 81.8a
its use, and efforts made to adjust for the the context of national surveillance, there 3 77.4 81.3a 85.1
underestimation of food insecurity. may be financial and space constraints 4 80.7a 84.4 89.0
The more comprehensive USDA-FSSM is associated with using the full set of 18-items, 5 83.4 86.5 92.9
available in three adaptations, which may and of note is the corresponding increase in 6 86.3 88.8 97.5
provide a more accurate picture of the burden missing data that accompanies the increase 7 89.2 92.1 NA
of food insecurity. Rasch analyses revealed in items included in each version of the scale. 8 90.6 93.5 NA
that overall the USDA-FSSM may be a valid The 6-item measure provided an estimate 9 92.7 94.5 NA
tool for the measurement of food insecurity of food insecurity similar to that provided 10 94.3 97.5 NA
within the current sample, with only a small by the 10 items, and two percentage points 11 95.9 NA NA
12 96.1 NA NA
number of items falling outside both the lower than that provided by the 18 items.
13 96.1 NA NA
standard and wider range for acceptable fit. Although this difference was statistically
14 96.3 NA NA
Across all three adaptations of the USDA- significant, compared with the single-item
15 96.3 NA NA
FSSM, the item assessing whether individuals from the NHS (which yielded a prevalence
16 96.3 NA NA
cut down the size of meals or skipped meals rate nearly five percentage points lower
17 96.5 NA NA
consistently yielded an INFIT value below 0.7, than that of the 18-item scale and was
18 96.9 NA NA
suggesting redundancy. The inclusion of this unable to identify different levels of food
Missing 3.1 2.5 2.5
item is unlikely to be problematic; instead, insecurity), the six items provided a better
Notes:
the information from this measure is likely representation of the prevalence and varying a: Meets prevalence of food insecurity as estimated by single-item NHS
to be slightly undervalued in assessing food levels of food insecurity. The 6-item scale measure
security status. The item investigating whether was also identified to be a valid measure of
individuals worried about running out of food food insecurity in the current sample, with In considering the USDA-FSSM as a tool for
before they could afford to purchase more only one question falling outside the wider the monitoring and surveillance of food
(included in the 18- and 10-item adaptations) accepted range for fit; this item, however, security, one notable limitation is the sole
yielded INFIT values above 1.3 (but below still falls within the range accepted as being focus on food insecurity arising from a lack
1.4), indicating more affirmative responses to productive to measurement, and thus of financial resources, meaning that this
this question than predicted by the model; exclusion at this stage is not warranted. scale is only able to assess a subcomponent
thus the information provided by this item In addition, the six items included in this of the overarching ‘access’ pillar (of a total
is potentially ‘overvalued’ slightly. In the six shorter iteration are similar to those included of four pillars underpinning food security:
child-related items contained within the in the Food Insecurity Experience Scale access, availability, utilisation and stability.1
18-item USDA-FSSM, the questions assessing (FIES). Recent work on the FIES has resulted As such, it is possible that the USDA-FSSM
the ability to feed children healthy meals and in the development of a process for the may also underestimate the true prevalence
the frequency with which children skipped calibration and subsequent comparison of of food insecurity. This may be particularly
meals also appeared to have been overvalued. other similar measures of food security on exacerbated among specific sub-groups of
Removal of items at this stage is unlikely to the basis of four or more items as anchoring the population, for example older and young
be warranted, given that all items fall within points, thus providing opportunity for adults, people living in isolation and/or rural
the range of 0.5–1.5, which indicates they international comparisons of food security and regional areas, and people with physical
are productive to the measurement of food data.31 Adoption of the 6-item version of the disabilities or specific health or dietary
insecurity,29 with the exception of the two USDA-FSSM in place of a single question in requirements. Future research efforts should
child-related items, which yielded INFIT values national surveillance efforts, such as the NHS, seek to identify ways to assess the remaining
of 1.61 and 1.90. However, the INFIT values of could therefore provide a better estimate pillars of food insecurity and accurately
these two items suggest that although these of the prevalence of food security, while capture its true burden on the population, as
values are not productive for the construction minimising cost and participant burden that indicated in recent Australian research.32,33
of a scale, they do not degrade the current would accompany using the complete 18- For further consideration is the potential for
measurement system. Further validation item scale, as well as provide opportunity for the full 18-item USDA-FSSM to underestimate
studies within a population-based sample, international comparisons. the prevalence of food insecurity among

2018 vol. 42 no. 4 Australian and New Zealand Journal of Public Health 393
© 2018 The Authors
McKechnie et al. Article

households without children when compared or physical disability and social isolation for measure, such as the 18-item USDA-FSSM,
specifically to households with children, when older adults, and poor food literacy for young which appears to be a valid measure for use
the simple summative method of scoring the adults), the survey tools used in the current among an Australian population and has
scale and classifying households according research would have been even less likely been shown to provide higher estimates
to standard cut-off points is used. This issue to adequately assess the prevalence of food of food insecurity and information on the
has been investigated previously by Nord,28 insecurity outside of the specific age range degree of severity of food security. In the
who concluded that to avoid this potential targeted in the current research. This is both context of financial and space constraints,
underestimation: a) the use and scoring of the a limitation of the current research, as well however, the 6-item USDA-FSSM may be a
adult-related items and child-related items as a limitation of existing tools to assess food more feasible option and has been shown to
should be performed separately, as described security status, and future research should provide an estimate of food insecurity closer
by Nord, and adopted in the current research; seek to develop comprehensive scales of to that of the 18-item tool.
b) food security status be classified using assessing multiple pillars of food insecurity
probabilistic assignment; or c) researchers across a broader age range. References
adopt clear and consistent acknowledgement In summary, the current method of assessing 1. Committee on World Food Security. Coming to Terms
throughout study findings of the bias and the food security status adopted in the three- with Terminology: Food Security, Nutrition Security, Food
reason it has occurred (this, however, is a less- Security and Nutrition, Food and Nutrition Security. Rome
yearly National Health Survey in Australia is (ITL): Food and Agricultural Organisation; 2012.
desirable approach as this information often likely underestimating the true prevalence 2. Kirkpatrick SI, Tarasuk V. Food insecurity is associated
fails to be translated in practice).28 of food insecurity. Compounding these with nutrient inadequacies among Canadian adults
and adolescents. J Nutr. 2008;138(3):604-12.
The findings of the current study should inaccuracies in measurement is the failure 3. Tarasuk VS. Household food insecurity with hunger
be interpreted within the context of the to consistently measure food security status is associated with women’s food intakes, health and
household circumstances. J Nutr. 2001;131(10):2670-6.
following limitations. First, sampling occurred in each subsequent National Health Survey 4. Wilde PE, Peterman JN. Individual weight change is
specifically within socioeconomically (with the single-question not included in associated with household food security status. J Nutr.
2006;136(5):1395 - 400.
disadvantaged urban areas of Brisbane, 2004–2005 or 2014–2015 data collection 5. Jones SJ, Frongillo EA. Food insecurity and
Australia. As such these findings may only efforts). The regular collection of accurate subsequent weight gain in women. Public Health
Nutr. 2007;10(2):145-51.
be generalisable to lower socioeconomic data pertaining to the prevalence of food 6. Stuff JE, Horton JA, Bogle ML, Connell C, Ryan D,
sub-groups of the population. To date, insecurity in any population is imperative to Zaghloul S, et al. High prevalence of food insecurity
and hunger in households in the rural Lower Mississippi
the only two known studies comparing assist researchers, practitioners and policy Delta. J Rural Health. 2004;20(2):173-80.
the USDA scales with the single Australian makers in ensuring equitable access to 7. Vozoris NT, Tarasuk VS. Household food insufficiency is
item have been undertaken in populations healthy, nutritious and culturally appropriate associated with poorer health. J Nutr. 2003;133(1):120-6.
8. Holben DH, Pheley AM. Diabetes risk and obesity in
residing in disadvantaged urban areas. foods for all people. The findings of the food-insecure households in rural Appalachian Ohio.
Further studies are required to determine current paper provide insight into the Prev Chronic Dis. 2006;3(3):1-9.
9. Heflin CM, Siefert K, Williams DR. Food insufficiency and
whether the differences identified between potential to include the 6-item iteration women’s mental health: Findings from a 3-year panel
these tools remain among a sample of of the USDA-FSSM in upcoming National of welfare recipients. Soc Sci Med. 2005;61(9):1971-82.
10. Alaimo K, Olson CM, Frongillo EA. Food insufficiency
the general population. In addition, our Health Surveys, and potentially in national and American school-aged children’s cognitive,
reliance on a 12-page mail-based survey surveillance efforts in other countries; the use academic, and psychosocial development. Pediatrics.
2001;108(1):44-53.
may have biased the sample against those of this shorter version would serve to limit 11. Cook JT, Frank DA, Berkowitz C, Black MM, Casey PH,
who live highly transient lifestyles, or who the financial and space constraints associated Cutts DB, et al. Food insecurity is associated with
are from non-English speaking backgrounds with the inclusion of larger measurements adverse health outcomes among human infants and
toddlers. J Nutr. 2004;134(6):1432-8.
and/or with poor literacy, resulting in scales, while providing opportunity for more 12. Kleinman RE, Murphy JM, Little M, Pagano M, Wehler
under-representation of some of the accurate assessment of the prevalence of CA, Regal K, et al. Hunger in Children in the United
States. Potential Behavioral and Emotional Correlates.
most disadvantaged segments of society, food insecurity (compared to the current Pediatrics. 1998;101(1):E3.
and those at highest risk of experiencing single-item), as well as international 13. Jyoti DF, Frongillo EA, Jones SJ. Food insecurity affects
school children’s academic performance, weight gain,
food insecurity. Finally, the recruitment of comparisons. However, the limitations of the and social skills. J Nutr. 2005;135(12):2831-9.
individuals aged between 20 and 59 years of tool only being able to assess one component 14. Klinteberg B, Schalling D, Manusson D. Childhood
behaviour and adult personality in male and female
age may to an extent limit the generalisability of the access pillar of food security should subjects. Eur J Pers. 1990;4:57-71.
of findings. This age range was selected continue to be acknowledged, and if possible 15. Fergusson D, Horwood L, Ridder E. Show me the child
for the purpose of sampling, and it is likely addressed by future research. at seven: The consequences of conduct problems in
childhood for psychosocial functioning in adulthood.
that people outside of this range were still J Child Psychol Psychiatry. 2005;46(8):837-89.
considered when answering questions 16. Bickel GW, Nord M, Price C, Hamilton WL, Cook J. Guide

that related to the household as a whole.

Conclusion to Measuring Household Food Security. Washington (DC):
United States America Department of Agriculture;
Nonetheless, it is possible that in specifying 2000.
Food insecurity is prevalent and a significant 17. Coleman-Jensen A, Rabbitt MP, Gregory CA, Singh A.
an age range for the purpose of recruitment, public health issue in disadvantaged Household Food Security in the United States in 2016
those who fell outside of this age group were urban areas of Australia. The current [Internet]. Washington (DC): United States Department
of Agriculture; 2017 [cited 2017 Nov 10]. Available
under-represented. However, given that single-item Australian measure potentially from: https://www.ers.usda.gov/publications/pub-
age groups above and beyond this specific underestimates the prevalence of food details/?pubid=84972
age range are more likely to experience 18. Coates J, Frongillo E, Rogers B, Webb P, Wilde P, Houser
insecurity. Screening and surveillance R. Advances in developing country food security
food insecurity for reasons beyond financial should seek to identify the prevalence of measurement. Commonalities in the experience of
access (for example limited mobility and/ household food insecurity across cultures: What are
food insecurity using a more comprehensive we missing? J Nutr. 2006;136:1338S–448S.

394 Australian and New Zealand Journal of Public Health 2018 vol. 42 no. 4
© 2018 The Authors
Data Collection Underestimation of food insecurity in Australia

19. Parnell W, Gray A. Development of a food security 25. Ramsey R, Giskes K, Turrell G, Gallegos D. Food insecurity 29. Wright B, Linacre J. Reasonable mean-square fit values.
measurement tool for New Zealand households. Br J among Australian children: Potential determinants, Rasch Meas Trans. 1994;8(30):370.
Nutr. 2014;112(8):1393-401. health and developmental consequences. J Child Health 30. Blumberg SJ, Bialostosky K, Hamilton WL, Briefel RR.
20. Australian Bureau of Statistics. National Health Survey Care. 2011;15(4):401-16. The effectiveness of a short form of the household food
Users’ Guide: 2004 - 2005. Canberra (AUST): ABS; 2006. 26. Ramsey R, Giskes K, Turrell G, Gallegos D. Food insecurity security scale. Am J Public Health. 1999;89(8):1231-4.
21. Australian Bureau of Statistics. 4363.0 - National among adults residing in disadvantaged urban areas: 31. Cafiero C, Viviani S, Nord M. Food security measurement
Health Survey 2014 – 2015: Questionnaire [Internet]. Potential health and dietary consequences. Public in a global context: The food insecurity experience
Canberra (AUST): ABS; 2017 [cited 2017 Nov 10]. Health Nutr. 2012;15(02):227-37. scale. Measurement. 2018;116:146-52.
Available from: http://www.abs.gov.au/ausstats/ 27. Australian Bureau of Statistics. Socio-economic 32. Archer C, Gallegos D, McKechnie R. Developing
abs@.nsf/Lookup/by%20Subject/~2014-15~Main%20 Indexes for Areas: Relative Measures of Socio-economic measures of food and nutrition security within an
Features~Users’%20Guide~1 Disadvantage and Advantage by Geographic Areas Australian context. Public Health Nutr. 2017;20(14):2513-
22. Australian Bureau of Statistics. Australian Health [Internet]. Canberra (AUST): ABS; 2011 [cited 2017 22.
Survey: Nutrition - State and Territory Results, 2011-12 Nov 10]. Available from: http://www.abs.gov.au/ 33. Kleve S, Gallegos D, Ashby S, Palermo C, McKechnie R.
[Internet]. Canberra (AUST): ABS; 2015 [cited 2017 websitedbs/censushome.nsf/home/seifa2011?open Preliminary validation and piloting of a comprehensive
Nov 10]. Available from: http://www.abs.gov.au/ document&navpos=260 measure of household food security in Australia. Public
ausstats/abs@.nsf/Lookup/4364.0.55.009main+featu 28. Nord M. Assessing Technical Enhancements to the U.S. Health Nutr. 2018;21(3):526-34.
res12011-12 Household Food Security Measures. Technical Bulletin
23. Nolan M, Rikard-Bell G, Mohsin M, Williams M. Food No.: 1936. Washington (DC): United States Department
insecurity in three socially disadvantaged localities in of Agriculture Economic Research Services; 2012.
Sydney, Australia. Health Promot J Austr. 2006;17:247-54.
24. Australian Bureau of Statistics. Census Dictionary.
Canberra (AUST): ABS; 2006.

2018 vol. 42 no. 4 Australian and New Zealand Journal of Public Health 395
© 2018 The Authors